scuffed-code/icu4c/source/i18n/csrucode.cpp
2017-01-20 00:20:31 +00:00

201 lines
5.4 KiB
C++

// © 2016 and later: Unicode, Inc. and others.
// License & terms of use: http://www.unicode.org/copyright.html
/*
**********************************************************************
* Copyright (C) 2005-2013, International Business Machines
* Corporation and others. All Rights Reserved.
**********************************************************************
*/
#include "unicode/utypes.h"
#if !UCONFIG_NO_CONVERSION
#include "csrucode.h"
#include "csmatch.h"
U_NAMESPACE_BEGIN
CharsetRecog_Unicode::~CharsetRecog_Unicode()
{
// nothing to do
}
CharsetRecog_UTF_16_BE::~CharsetRecog_UTF_16_BE()
{
// nothing to do
}
const char *CharsetRecog_UTF_16_BE::getName() const
{
return "UTF-16BE";
}
// UTF-16 confidence calculation. Very simple minded, but better than nothing.
// Any 8 bit non-control characters bump the confidence up. These have a zero high byte,
// and are very likely to be UTF-16, although they could also be part of a UTF-32 code.
// NULs are a contra-indication, they will appear commonly if the actual encoding is UTF-32.
// NULs should be rare in actual text.
static int32_t adjustConfidence(UChar codeUnit, int32_t confidence) {
if (codeUnit == 0) {
confidence -= 10;
} else if ((codeUnit >= 0x20 && codeUnit <= 0xff) || codeUnit == 0x0a) {
confidence += 10;
}
if (confidence < 0) {
confidence = 0;
} else if (confidence > 100) {
confidence = 100;
}
return confidence;
}
UBool CharsetRecog_UTF_16_BE::match(InputText* textIn, CharsetMatch *results) const
{
const uint8_t *input = textIn->fRawInput;
int32_t confidence = 10;
int32_t length = textIn->fRawLength;
int32_t bytesToCheck = (length > 30) ? 30 : length;
for (int32_t charIndex=0; charIndex<bytesToCheck-1; charIndex+=2) {
UChar codeUnit = (input[charIndex] << 8) | input[charIndex + 1];
if (charIndex == 0 && codeUnit == 0xFEFF) {
confidence = 100;
break;
}
confidence = adjustConfidence(codeUnit, confidence);
if (confidence == 0 || confidence == 100) {
break;
}
}
if (bytesToCheck < 4 && confidence < 100) {
confidence = 0;
}
results->set(textIn, this, confidence);
return (confidence > 0);
}
CharsetRecog_UTF_16_LE::~CharsetRecog_UTF_16_LE()
{
// nothing to do
}
const char *CharsetRecog_UTF_16_LE::getName() const
{
return "UTF-16LE";
}
UBool CharsetRecog_UTF_16_LE::match(InputText* textIn, CharsetMatch *results) const
{
const uint8_t *input = textIn->fRawInput;
int32_t confidence = 10;
int32_t length = textIn->fRawLength;
int32_t bytesToCheck = (length > 30) ? 30 : length;
for (int32_t charIndex=0; charIndex<bytesToCheck-1; charIndex+=2) {
UChar codeUnit = input[charIndex] | (input[charIndex + 1] << 8);
if (charIndex == 0 && codeUnit == 0xFEFF) {
confidence = 100; // UTF-16 BOM
if (length >= 4 && input[2] == 0 && input[3] == 0) {
confidence = 0; // UTF-32 BOM
}
break;
}
confidence = adjustConfidence(codeUnit, confidence);
if (confidence == 0 || confidence == 100) {
break;
}
}
if (bytesToCheck < 4 && confidence < 100) {
confidence = 0;
}
results->set(textIn, this, confidence);
return (confidence > 0);
}
CharsetRecog_UTF_32::~CharsetRecog_UTF_32()
{
// nothing to do
}
UBool CharsetRecog_UTF_32::match(InputText* textIn, CharsetMatch *results) const
{
const uint8_t *input = textIn->fRawInput;
int32_t limit = (textIn->fRawLength / 4) * 4;
int32_t numValid = 0;
int32_t numInvalid = 0;
bool hasBOM = FALSE;
int32_t confidence = 0;
if (limit > 0 && getChar(input, 0) == 0x0000FEFFUL) {
hasBOM = TRUE;
}
for(int32_t i = 0; i < limit; i += 4) {
int32_t ch = getChar(input, i);
if (ch < 0 || ch >= 0x10FFFF || (ch >= 0xD800 && ch <= 0xDFFF)) {
numInvalid += 1;
} else {
numValid += 1;
}
}
// Cook up some sort of confidence score, based on presense of a BOM
// and the existence of valid and/or invalid multi-byte sequences.
if (hasBOM && numInvalid==0) {
confidence = 100;
} else if (hasBOM && numValid > numInvalid*10) {
confidence = 80;
} else if (numValid > 3 && numInvalid == 0) {
confidence = 100;
} else if (numValid > 0 && numInvalid == 0) {
confidence = 80;
} else if (numValid > numInvalid*10) {
// Probably corruput UTF-32BE data. Valid sequences aren't likely by chance.
confidence = 25;
}
results->set(textIn, this, confidence);
return (confidence > 0);
}
CharsetRecog_UTF_32_BE::~CharsetRecog_UTF_32_BE()
{
// nothing to do
}
const char *CharsetRecog_UTF_32_BE::getName() const
{
return "UTF-32BE";
}
int32_t CharsetRecog_UTF_32_BE::getChar(const uint8_t *input, int32_t index) const
{
return input[index + 0] << 24 | input[index + 1] << 16 |
input[index + 2] << 8 | input[index + 3];
}
CharsetRecog_UTF_32_LE::~CharsetRecog_UTF_32_LE()
{
// nothing to do
}
const char *CharsetRecog_UTF_32_LE::getName() const
{
return "UTF-32LE";
}
int32_t CharsetRecog_UTF_32_LE::getChar(const uint8_t *input, int32_t index) const
{
return input[index + 3] << 24 | input[index + 2] << 16 |
input[index + 1] << 8 | input[index + 0];
}
U_NAMESPACE_END
#endif